Fast Magic-Angle Spinning Three-Dimensional NMR Experiment for Simultaneously Probing H  H and N  H Proximities in Solids

: A fast magic-angle spinning (MAS, 70 kHz) solid-state NMR experiment is presented that combines 1 H Double-Quantum (DQ) and 14 N  1 H HMQC (Heteronu-clear Multiple-Quantum Coherence) pulse-sequence elements, so as to simultaneously probe H  H and N  H proximities in molecular solids. The proposed experiment can be employed in both two-dimensional (2D) and three-dimensional (3D) versions: ﬁ rst, a 2D 14 N HMQC- ﬁ ltered 1 H-DQ experiment provides speci ﬁ c DQ-SQ correlation peaks for proton pairs that are in close proximities to the nitrogen sites, thereby achieving spectral ﬁ ltration. Second, a proton-detected three-dimensional (3D) 1 H(DQ)- 14 N(SQ)- 1 H(SQ) experiment correlates 1 H(DQ)- 1 H(SQ) chemical shifts with 14 N shifts such that longer range N ··· H  H correlations are observed between protons and nitrogen atoms with internuclear NH distances exceeding 3 Å. Both 2D and 3D versions of the proposed experiment are demonstrated for an amino acid hydrochloride salt, L -histidine · HCl · H 2 O, and a DNA nucleoside, guanosine · 2H 2 O. In the latter case, the achieved spectral ﬁ ltration ensures that DQ cross peaks are only observed for guanine NH and CH8 1 H resonances and not ribose and water 1 H resonances, thus providing insight into the changes in the solid-state structure of this hydrate that occur over time; signi ﬁ cant changes are observed in the NH and NH 2 1 H chemical shifts as compared to the freshly recrystallized sample previously studied by Reddy et al., Cryst. Growth Des. 2015 , 15 , 5945. using BaBa-xy16 recoupling and 14 N  1 H HMQC spectroscopy using SR4 recoupling pulse sequence elements. Both the 1 H-DQ evolution during t 1 and the 14 N evolution period during t 2 are rotor- synchronized with respect to the sample spinning. In the case of a 2D 1 H  14 N HMQC- ﬁ ltered 1 H DQ-SQ correlation experiment, only the t 1 period is incremented and t 2 is set equal to one rotor period.


■ INTRODUCTION
Within the context of small and moderately sized organic molecules, 1 H solid-state NMR is finding increasing application for the solid-state characterization of a range of samples including pharmaceuticals, supramolecular assemblies and polymers, 1 primarily benefiting from its inherently high sensitivity even at submilligram quantities, for example, using fast magic-angle spinning (MAS) approaches. 2,3 Notably, the 1 H chemical shift is sensitive to intermolecular interactions such as hydrogen bonding and aromatic π−π interactions, which govern molecular packing in the solid-state. 4−12 Moreover, 1 H detection is becoming increasingly important in biological solid-state NMR. 13−17 Solid-state 1 H MAS NMR spectra of organic molecules often exhibit markedly overlapped resonances; nevertheless, 2D homonuclear 1 H Double-Quantum 18 (for example recorded using BaBa 19−21 recoupling), NOESY-like spin diffusion, 22, 23 and heteronuclear 1 H 13 C and 15 N/ 14 N 1 H correlation experiments significantly aid spectral interpretation. Specifically, NH proximities can be probed using a 14 N 1 H HMQC experiment 3,24−27 ( 14 N spin 1, 99.6% natural isotopic abundance) capitalizing on J couplings, dipolar couplings, and also recently on overtone 14 N transitions. 28−30 The 14 N 1 H HMQC experiment has been employed to characterize intermolecular NH···N, NH···O and OH···N hydrogen bonding interactions in pharmaceutical cocrystals 31,32 and in the self-assembled structures adopted by modified DNA/RNA nucleosides. 33−36 Here, a three-dimensional 1 H(DQ)-14 N(SQ)-1 H(SQ) correlation experiment (Figure 1) is presented whereby, under fast MAS (70 kHz), 1 H DQ-SQ and 1 H 14 N HMQC pulse sequence elements are concatenated. This experiment sequentially utilizes both HH and NH dipolar interactions for probing N···HH proximities, and has the following advantages: first, incorporation of a 14 N-HMQC filter in the 1 H DQ experiment achieves spectral filtration by retaining only specific DQ correlation peaks for protons which are in a close proximity to nitrogen sites. This can be achieved simply by recording a 2D version of the experiment. It is to be noted that a similar approach has been proposed by Spiess and co-workers for achieving spectral simplification via 15 N-edited 1 H-DQ MAS NMR spectroscopy in keto and enol tautomers of 15 N-labeled N-butylaminocarbonyl-6-tridecyl isocytosine. 37 Second, in a three-dimensional 1 H(DQ)-14 N(SQ)-1 H(SQ) experiment, proton pairs that are remotely dipolar coupled with 14 N sites (in which the interatomic distance between 14 N-sites and one of the protons in the HH pair exceeds 3 Å) are observed. Both 2D and 3D versions of the 1 H(DQ)-14 N(SQ)-1 H(SQ) correlation experiment are demonstrated here for probing HH and NH proximities in an amino acid hydrochloride salt, L-histidine·HCl·H 2 O and a dihydrate of the nucleoside guanosine, G·2H 2 O. Our aim here is, thus, to show the applicability of the MAS NMR experiments to such small and moderately sized organic molecules.

■ EXPERIMENTAL AND COMPUTATIONAL DETAILS
L-histidine·HCl·H 2 O and G·2H 2 O were purchased from Sigma-Aldrich, Gillingham, U.K. Guanosine was recrystallized from water according to the procedure described in ref 35 and stored under standard laboratory conditions. Powder X-ray Diffraction (PXRD). Data were collected at room temperature on a Bruker D8 Advance (Kα 1 λ = 1.5406 Å) equipped with monochromatic Cu K α1 radiation and a Nifiltered VÅNTEC1 detector (L-Histidine·HCl·H 2 O), and PANalytical X′Pert Pro MPD (Kα 1 λ = 1.5406 Å) equipped with monochromatic Cu K α1 radiation and a PIXcel detector (G·2H 2 O). It was verified that the samples corresponded to the HISTCM01 38  H and 14 N pulse lengths were experimentally optimized using a one-dimensional version of the pulse sequence shown in Figure 1. The 1 H π/2 pulse duration was 0.92 μs in the HMQC pulse sequence element, while a 1 H (nominal) π/2 pulse duration of 0.5 μs (with the same rf nutation frequency) was used in the BaBa-xy16 block. The 14 N pulse duration was 10 μs. 1 H One-Pulse. Sixteen transients were coadded using a recycle delay of 6 s (L-histidine·HCl·H 2 O) or 3 s (G·2H 2 O). 13 C CPMAS. 1D 13 C CPMAS experiments were performed on a Bruker Avance III 500 MHz ( 13 C, 125 MHz) spectrometer equipped with a 4.0 mm triple resonance MAS probe (operating in double resonance mode). For the sample stored 1 year after recrystallization, 42 mg of G.2H 2 O was packed into a 4 mm (outer diameter) rotor. 1 H and 13 C π/2 pulse lengths were 2.5 and 3.5 μs, respectively. The cross-polarization contact time was 1.5 ms. Heteronuclear decoupling using SPINAL64 40 with a 1 H pulse duration of 5 μs was applied for an acquisition time of 30 ms. 2048 transients were coadded using a relaxation delay of 3 s for a total experimental time of 2 h. 1 H Double-Quantum (DQ) Spectroscopy. Eight rotor periods (corresponding to 114 μs) of BaBa-xy16 recoupling was used for the excitation and reconversion of DQ coherencesthe BaBa-xy16 sequence consists of a train of 32 π/2 pulses synchronized to eight rotor periods. 21 A 16-step phase cycle was used in order to select a change in coherence order Δp = ± 2 on the DQ excitation pulses (4 steps) and Δp = −1 (4 steps) on the z-filter 90°pulse, where p is the coherence order. For both L-histidine·HCl·H 2 O and G·2H 2 O, 32 t 1 FIDs, each with 16 coadded transients, were acquired using the States-TPPI method to achieve sign discrimination in the F 1 dimension with a rotor-synchronized t 1 increment of 14.3 μs. The total experimental times was 1.7 h using a 6 s recycle delay for L-histidine·HCl·H 2 O and 45 min for G.2H 2 O using a 3 s relaxation delay. A BaBa-xy16 sequence 21 is used for the excitation of DQ coherence and reconversion of DQ into SQ coherence, with the resulting DQ-filtered SQ coherence being the starting point for a subsequent 14 N 1 H HMQC pulse sequence element. The BaBa-xy16 part is phase cycled to observe DQ coherences and the HMQC part is phase cycled to select the SQ coherence pathways. In the HMQC part, the SR4 pulse sequence was applied during the excitation and reconversion periods to decouple the 1 H homonuclear dipolar interactions and recouple the 1 H 14 N heteronuclear dipolar interactions. 41 For the pulse sequence shown in Figure  1, the phase cycling is given as follows:  Pulse sequence for a three-dimensional 1 H(DQ)-14 N-(SQ)-1 H(SQ) correlation experiment that combines 1 H(DQ) spectroscopy using BaBa-xy16 recoupling and 14 N 1 H HMQC spectroscopy using SR4 recoupling pulse sequence elements. Both the 1 H-DQ evolution during t 1 and the 14 N evolution period during t 2 are rotorsynchronized with respect to the sample spinning. In the case of a 2D 2(300, 120, 120, 300), 2(120, 300, 300, 120), and 2(300, 120, 120, 300)]. Both the t 1 duration for 1 H DQ evolution and the t 2 duration for 14 N evolution are rotor synchronized with the sample spinning, i.e., t 1 = mτ r and t 2 = nτ r , where m and n are integers.
In the case of the 2D version of the 1 H(DQ)-14 N(SQ)-1 H-(SQ) experiment, only the t 1 period was incremented. A total of 24 t 1 FIDs were acquired using the States-TPPI method to achieve sign discrimination in F 1 with a rotor synchronized increment of 14.3 μs. For L-histidine·HCl·H 2 O, 96 transients were coadded with a recycle delay of 6 s corresponding to a total experimental time of 7.7 h, while for G·2H 2 O, 32 transients were coadded with a recycle delay of 3 s corresponding to a total experimental time of about 5.5 h.
In the case of 3D experiments, an array of 16 t 1 × 16 t 2 FIDs were collected in the 1 H DQ and 14 N SQ dimensions using the States-TPPI method. For L-histidine·HCl·H 2 O, 32 transients were coadded with a recycle delay of 5 s corresponding to a total experimental time of 45.5 h., while for G·2H 2 O, 96 transients were coadded with a recycle delay of 3 s corresponding to a total experimental time of 82 h.
All 1 H chemical shifts are calibrated with respect to neat TMS using adamantane (1.85 ppm) as an external reference. 42 14 N chemical shifts were referenced to neat CH 3 NO 2 using powdered NH 4 Cl at −341.2 as an external reference (see Table  2 of ref 43). To convert to the chemical shift scale frequency used in protein NMR, where the alternative IUPAC (see Appendix 1 of ref 44) reference is liquid NH 3 at −50°C, it is necessary to add 379.5 ppm to the given values. 45 GIPAW DFT Calculations. All calculations were performed using plane-wave based DFT implemented within the Cambridge Serial Total Energy Package (CASTEP) code, U.K. academic release version 8.0. 46 Atomic coordinates were obtained from the crystal structure of L-histidine.HCl.H 2 O, as previously solved by X-ray diffraction: CSID code HISTCM01, Z = 4, Z′ = 1, space group P2 1 , 100 atoms/unit cell (including 1 HCl and 1 H 2 O). 38 In a first stage, a geometry optimization is performed: starting with the crystal structure, the positions of all atoms are allowed to move (with the unit cell parameters fixed, and space group symmetry imposed as determined from the X-ray diffraction structure) until an energy-minimized structure is obtained. The distances stated in this paper correspond to this geometry-optimized crystal structure. NMR shielding calculations were performed using the Gauge-Including Projector-Augmented Wave (GIPAW) approach. 47,48 Both geometry optimization and NMR chemical shift calculations used a plane-wave basis set and the PBE exchange correlation functional 48,49 at a basis cutoff energy of 600 eV with integrals taken over the Brillouin zone by using a Monkhorst−Pack grid of minimum sample spacing 0.08 × 2π Å −1 . A semiempirical dispersion correction was applied using the TS scheme 50 for both geometry optimization and NMR shielding calculations with on-the-fly (OTF) ultrasoft pseudopotentials. 51 Forces, stress on the unit cell, energy and displacements were converged to better than 0.01 eV Å −1 , 0.1 G Pa, 0.000 000 4 eV, and 0.001 Å, respectively.
■ RESULTS AND DISCUSSION L-Histidine·HCl·H 2 O. A single-pulse 1 H (700 MHz, 70 kHz MAS) spectrum is presented in Figure 2a: the higher ppm peaks assigned to H5 and H7 correspond to protons exhibiting intermolecular NH···O hydrogen bonding interactions.  Table S1, while GIPAW DFT calculated 1 H chemical shielding values are given in Table S2. Rows extracted from the regular 2D 1 H DQ and the 14 N-HMQC filtered 1 H DQ spectra are presented for the DQ peaks at 5.8, 10.3, 16.0, and 29.4 ppm in Figure S1.
In a 14 N-HMQC filtered 1 H DQ-SQ spectrum recorded using a short recoupling time (114 μs, Figure 2c), DQ-SQ peaks are only retained for proton pairs where one or both

Analytical Chemistry
Article protons are directly bonded to a 14 N site, i.e., for the ring NH (H5 and H7) and NH 3 (H1) protons. Note that low intensity for the NH 3 (H1) peak is attributed to the rotation of the NH 3 protons reducing the magnitude of the 14 N 1 H dipolar coupling by which magnetization transfer is achieved. In the 14 N-filtered DQ spectrum recorded by using a longer recoupling duration (286 μs, Figure 2d), 1 H DQ-SQ peaks are recovered for the NH 3 (H1) peak and the imidazole ring protons (H6 and H8) for which the nearest N···H distances are between 2.15 and 2.17 Å. It is evident from Figure 2c,d that the spectral complexity in the 1 H DQ spectra can be readily tuned by varying the recoupling duration, thus aiding spectral interpretation.
The left-hand side of Figure 3 presents 14 N(SQ)-1 H(SQ) spectra recorded with the same recoupling durations as for the 14 N-HMQC filtered 1 H DQ-SQ spectra presented in Figure  2c,d, namely a short 114 μs (Figure 3a) and a longer 286 μs (Figure 3b) recoupling duration. As was the case for the spectra in Figure 2, by changing the recoupling duration, the spectroscopist can select the number of correlation peaks that     Figure 2d (for the longer recoupling duration). Note that Table S3 lists calculated (GIPAW) 14 N chemical shieldings and quadrupolar parameters for L-histidine·HCl·H 2 O and compares the calculated and experimental 14 N shifts. Figure 3 also presents spectral planes extracted from two different 1 H(DQ)-14 N(SQ)-1 H(SQ) HMQC three-dimensional experimental data sets. Such a 3D experiment (see Figure 1) permits the sampling, in the first and second indirect dimensions (i.e., t 1 and t 2 ), of both 1 H DQ and 14 N(SQ) coherences, while benefiting from the advantage of 1 H detection 52−55 (i.e., during the acquisition period, t 3 ). From the cuboid resulting from a three-dimensional Fourier transformation, it is possible to extract planes corresponding to two of the three spectral dimensions at a fixed frequency in the third dimension. As shown in Figure 3c and 3d, it is of most interest to consider 14 N(SQ)-1 H(DQ) planes extracted at the 1 H SQ chemical shifts corresponding to the three directly bonded NH moieties for which cross peaks are observed in Figure 3a, i.e., at 17.0 ppm (H1, NH), 12.4 ppm (H7, NH) and 8.0 ppm (H1, NH 3 ). The planes presented in Figure 3c and 3d correspond to two different 3D experiments using the two distinct recoupling times of 114 and 286 μs, respectively.
The most intense peaks in the planes shown in Figure 3c, It is also important to observe that the 14 N(SQ)-1 H(DQ) planes at the H5 (17 ppm) 1 H chemical shift display a N5··· H5H7 correlation peak and likewise the 14 N(SQ)-1 H(DQ) planes at the H7 chemical shift (12.4 ppm) display a N7··· H5H7 correlation peak. Note that the closest HH proximity for this H5H7 DQ coherence is an intermolecular proximity of 3.19 Å, with the intramolecular proximity being 4.12 Å. In such a case, the contribution of multiple different HH distances is reflected in a root-sum squared dipolar coupling, with the intensity of a 1 H DQ peak depending, to a first approximation, on the square of this root-sum squared dipolar coupling, 56,57 i.e., the contribution of the closer intermolecular proximity compared to the intramolecular proximity to the H5H7 DQ can be estimated as 4.12 6 / 3.19 6 ∼ 5. As listed in Table 1 and illustrated by the inset in the bottom left of Figure 3, the corresponding N5···H7 and N7··· H5 distances are 3.44 and 3.84 Å (intermolecular) and 3.19 and 3.15 Å (intramolecular), which is significantly longer than the under 2.2 Å longer-range nitrogen hydrogen distances for the cross peaks observed in the 2D 14 N(SQ)-1 H(SQ) spectrum in Figure 3b.
Guanosine·2H 2 O. Supramolecular assemblies generated from guanosine (G) derivatives have a wide range of applications such as lyotropic mesophases, gelators, thin-films, and synthetic ion channels. 58,59 The formation of ribbon-like assemblies in the absence of cations, mainly driven by intermolecular NH···N and NH···O hydrogen bonding interactions, is well-known. 34,35,39,60,61 A crystal structure of guanosine, G·2H 2 O, has been solved by X-ray diffraction; there is a ribbon-like assembly whereby two crystallographically independent molecules, namely A and B, self-organize in the form of -A-A-A-A-and -B-B-B-B-. 39 In this structure, there are two types of water molecules: interlayer water (W1 and W2 interconnect the sugar moieties of adjacent G-ribbons) and intralayer water (W3 and W4 interconnecting the twodimensional sheets of guanine frames) that reinforce the three-dimensional stacking of ribbons by means of intermolecular O−H···O hydrogen bonds ( Figure S4). 35,39,62,63 Figure 4a reproduces from ref 35 (Figures 3 and S2) a 1D 1 H and 2D 1 H DQ-SQ MAS spectrum of G·2H 2 O as recorded directly after the recrystallization from water. As shown in Figure S4, it was verified by PXRD that this sample corresponds to the published crystal structure, GUANSH10. 39 For this recrystallized sample, after storage for a year under laboratory conditions, 1D 1 H and 2D 1 H DQ-SQ MAS spectra were recorded again as shown in Figure 4b. There is an evident change in the spectra, with this being consistent with the PXRD pattern (see Figure S4) having also changed considerably. Notably, the 1 H chemical shifts of the NH1 and NH 2 protons Sugawara and co-workers have observed changes in PXRD patterns for the solid-state structures of guanosine associated with changing moisture content; using molecular dynamics (MD), they have identified that the intralayer water molecules (W3 and W4) have a relatively weak affinity to the G-ribbons, hence offering an explanation for the humidity-induced changes in solid-state structure. 62,63 Figure 4c compares thermogravimetric analysis (TGA) of G·2H 2 O, carried out directly after the recrystallization from water (black dashed lines, also presented in Figure 2 of ref 35) and performed a year later after the sample had been stored at laboratory conditions (green solid line). Both samples showed 12.5% weight loss corresponding to the dihydrate form of guanosine, but minor changes are observed in the TGA curves in the temperature range 40°C to 120°C, suggesting variations in affinity of water molecules within the crystal lattice that are in line with the experimental observations and MD simulations of Sugawara and coworkers. 62,63 We show here the structural insight that can be gained from the new solid-state NMR experiment described above for the case of this guanosine dihydrate sample (stored under laboratory conditions for a year after recrystallization) for which there is no crystal structure available. In addition, a comparison of 13 C cross-polarization (CP) MAS spectra of an as-received guanosine sample, the fresh recrystallized sample and the sample in this work is given in Figure S4; the corresponding 13 C chemical shifts are stated in Table S4, noting that Sugawara et al. have presented 13 C CPMAS spectra reported in Figure 3 of ref 62. 2D 14 N-HMQC filtered 1 H DQ-SQ correlation spectra recorded by using a short (114 μs) and a long (571 μs) recoupling time are presented in Figure 5b,c, respectively; for comparison, a standard 1 H DQ spectrum presented in Figure   5a is repeated from Figure 4b above. Specific rows extracted from the 2D 14 N-HMQC filtered 1 H DQ-SQ correlation spectra are presented in Figure S5. The advantage of the spectral filtration achieved in this case for the crowded spectral region for δ SQ between 3 and 9 ppm is evident. Specific DQ cross peaks become visible that are otherwise obscured by the broad ribose and water 1 H SQ resonances.
In order to assign the observed 1 H DQ peaks in Figure 5, consider first the complementary insight that is provided by the two-dimensional 14 N 1 H spectra in Figure 6a,b that were recorded with the same short (114 μs) and a long (571 μs) recoupling time. In Figure 6a, cross peaks corresponding to the one-bond NH connectivities in the NH (NH1) and NH 2 (NH2a and NH2b) moieties are observed, while in Figure 6b, cross peaks corresponding to longer range N···H proximities involving the H8 and sugar protons (H1′) and the nonprotonated nitrogen resonances (N7, N9) appear. Returning to Figure 5, it can then be identified that in Figure 5b (τ rcpl = 114 μs), DQ-SQ peaks are largely only observed for the NH (H1) and NH 2 (NH2a and NH2b) 1 H SQ resonances, while for the spectrum in Figure 5c (τ rcpl = 571 μs), DQ peaks are additionally seen for the CH8 1 H SQ resonance that has a close (<2.2 Å) intramolecular NH proximity. Importantly, this allows the assignment of 1 H DQ cross peaks at 4.9 + 13.3 = 18.2 ppm between NH (H1) and the NH2b protons in Figure 5b and at 13.3 + 7.8 = 21.1 ppm between NH (H1) and the H8 protons in Figure 5c. As shown for other guanosine derivatives, the changes in SQ and DQ frequencies for such peaks can be structurally informative about different modes of selfassembly. [33][34][35]64 As was the case with L-histidine·HCl·H 2 O above, further insight is provided by considering Figure 6c   and NH (H1) protons, respectively. The planes in Figure 6c correspond to a short (114 μs) duration of recoupling of the 14 N 1 H dipolar interaction such that, as in Figure 6a, strongest intensity is observed at the N1 14 N shift for the plane at the NH (H1) 1 H SQ chemical shift. In this case, intense peaks are observed at the H1 + H8 and H1 + H2b 1 H DQ chemical shifts as in Figure 5b, corresponding to N1···H8 and N1···H2b longer-range proximities. For the planes in Figure 6d corresponding to a long (571 μs) recoupling duration, peaks are also observed at the other nitrogen resonances: consider the plane at the H8 1 H SQ chemical shift, strong intensity is observed for 1 H DQ peaks involving the H8 proton and the N7 and also the N1 nitrogen. Such information about specific longer-range nitrogen hydrogen distances could be useful for testing structural models for this guanosine dihydrate structure, e.g., those deriving from a MD simulation such as in ref 63.

■ CONCLUSIONS
This study presents 2D and 3D versions of a 1 H(DQ)-14 N-(SQ)-1 H(SQ) NMR experiment for probing simultaneously HH proximities and NH proximities under fast MAS conditions. Experimental results are demonstrated for an amino acid salt, L-histidine·HCl·H 2 O and a DNA base, G·2H 2 O. 2D 14 N-edited 1 H DQ-SQ spectra retain specific DQ peaks for proton-pairs that are dipolar coupled with 14 N sites. This type of spectral filtration led to the observation of specific DQ peaks involving the H8 and NH2 protons in G·2H 2 O, which are otherwise lost under DQ peaks due to ribose and water protons in a standard 1 H DQ-SQ correlation spectrum. Furthermore, a 3D 1 H(DQ)-14 N(SQ)-1 H(SQ) correlation experiment allows the observing of long-range 14 N 1 H correlations, as shown for N···H distances exceeding 3 Å for L-histidine·HCl·H 2 O.
The 1 H(DQ)-14 N(SQ)-1 H(SQ) experiment is complementary to previously reported 15 N-filtered 1 H DQ-SQ 37 ( 15 N isotopic labeling was required) experiments as well as 1 H DQ-13 C SQ or 14 N-filtered 1 H 13 C correlation experiments (larger sample quantities are required). 65−67 The proposed experiment is particularly useful for obtaining specific structural information or monitoring temporal changes in the vicinity of NH sites. This approach could be further extended to probe HH proximities in the vicinity of other NMR active quadrupolar nuclei, e.g., chlorine 68 which will also benefit from the use of higher magnetic field strength under fast MAS conditions. ■ ASSOCIATED CONTENT

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.6b01869.